The prior work in implicit had some specific assumptions about the workload modeling and the kind of environments in which the jobs were to be executed. They had assumed a bulk-synchronous SPMD model of parallel jobs, where phases of computations and communication were clearly separable. Also, they had used a fixed and arbitrary context switch time for all the processes, which did not properly reflect the actual overhead of the workloads being simulated.
In this project, we focussed on more generic applications, namely Barnes-Hut, Water and Sor, and compared the performance of the two techniques with various combinations of these jobs competing for the limited processing resource.
Also, we used a better estimate of the context switch time for each process to better decide on the period over which a process should spin wait. A more dynamic scheme of process context switch times which varies over the execution lifetimes of the processes might be more helpful in making the decisions.